Implementing Jax backend, migrating TensorFlow and PyTorch backends to qiboml

[MatteoRobbiati]

This is more a notification PR of the ongoing work.
A primitive version of the pytorch backend is now implemented in qiboml (I locally tested it in Qibo).
With the new qiboml, the idea is to lighten Qibo from the quantum machine learning backends.

Checklist:

• Reviewers confirm new code works as expected.
• Tests are passing.
• Coverage does not decrease.
• Documentation is updated.